Re: [PATCH v6 06/11] x86/hvm: processor trace interface in HVM

[Jan Beulich <jbeulich@xxxxxxxx>]

On 07.07.2020 21:39, Michał Leszczyński wrote:
> --- a/xen/arch/x86/domain.c
> +++ b/xen/arch/x86/domain.c
> @@ -2205,6 +2205,27 @@ int domain_relinquish_resources(struct domain *d)
>                  altp2m_vcpu_disable_ve(v);
>          }
>  
> +        for_each_vcpu ( d, v )
> +        {
> +            unsigned int i;
> +            uint64_t nr_pages = v->domain->processor_trace_buf_kb * KB(1);
> +            nr_pages >>= PAGE_SHIFT;
> +
> +            if ( !v->vmtrace.pt_buf )
> +                continue;
> +
> +            for ( i = 0; i < nr_pages; i++ )
> +            {
> +                struct page_info *pg = mfn_to_page(
> +                    mfn_add(page_to_mfn(v->vmtrace.pt_buf), i));
> +
> +                put_page_alloc_ref(pg);
> +                put_page_and_type(pg);
> +            }
> +
> +            v->vmtrace.pt_buf = NULL;
> +        }

This needs to allow for preemption. Also this isn't x86-specific,
so should be implemented in common code (just like allocation is).

Jan